有限记忆量测噪声在线估计的Kalman改进算法

Improved algorithm for limited memory on-line estimating Kalman filter

针对有限记忆量测噪声在线估计算法中,新息残差序列对渐变噪声的统计滞后问题,提出了一种新息变化率构建算法.该算法利用当前统计周期内新息绝对值的均值与前一统计周期内新息均值的绝对值构造新息变化率,并在此基础上提出了利用新息变化率作为量测噪声估计阵修正因子的改进算法.仿真比较了在无噪声变化、噪声突变与噪声渐变3种不同情况下,算法改进前后的滤波效果.仿真结果表明,该算法在保留对突变噪声有效检测的同时,提高了对渐变噪声的检测速度,从而提高了有限记忆量测噪声在线估计算法对量测噪声阵的计算精度.

For the lag of innovation residual sequence to gradually changing noise in the limited memory filter,a changing rate construction algorithm is proposed,in which the mean of absolute innovation value in the current statistical period and the absolute mean of innovation value in the last statistical period are used to construct the innovation changing rate.An improved algorithm is presented in which the innovation changing rate is used to correct the estimated matrix of measurement noise.The filter results before and after the algorithm improvement are compared in three different conditions： no noise changing,abrupt noise changing and gradual noise changing.And the simulation results show that with the algorithm proposed the reaction rate to gradual changing noise is increased.